Paul and Sabrina's Cheap 3 Phase Inverter (AC Controller) with Field Oriented Control

[thingstodo]

Originally Posted by MPaulHolmes

It took about 7 hours, all by hand. The surface mount part isnt too bad. The smallest sm parts are 1210.

I've tested the 6 isolated supplies. So far so good.

WOW - 7 hours!?

My eyes go fuzzy after about an hour of detailed work (not soldering, though. I haven't soldered for more than 5 minutes in a row for years)

Great to hear testing is going well.

[MPaulHolmes]

Well, that was also carefully verifying the value of each part in the schematic though. The software is ready to start testing/debugging on the bench without the motor.

[4x4EV]

Fall City, Wa It is between issaquah and northbend washington in the cascade mountains

[4x4EV]

this guys has two of these combos available, shipping is unknown and looks like they would need to get taken apart and atleast cleaned, but its still thousands off retail

AC Curtis Motor Controller Motor Combo | eBay

[e*clipse]

Very nice work on that circuit board Paul!

Yes, the surface mount stuff isn't too big a deal to solder. In fact some of the bigger stuff really isn't any different. I've also found that they can help with certain performance issues. As long as one stays away from the really tiny stuff, they open up quite a few options.

- I'd better get to work!!

- E*clipse

Originally Posted by MPaulHolmes

Top:

Bottom:

[thingstodo]

Originally Posted by MPaulHolmes

Well, that was also carefully verifying the value of each part in the schematic though. The software is ready to start testing/debugging on the bench without the motor.

The parts verify would be more 'eyes go fuzzy' work if I were to do it.

I can't remember if I posted this - a plan for testing the AC controller:
- a netgain 9 inch DC motor with an existing 1000A 400V controller
- my 30 HP AC motor with Paul's AC controller
- coupled with the largest joy coupler that I can get to fit
- set up the AC controller for regen
- both controllers run from same battery pack - 400V
- battery charger set up to 'top up' battery pack as energy is removed
- run up the AC motor with the AC controller
- turn up the DC controller until the rpm begins to increase (AC current drops off)
- creep down the speed of the AC controller and begin regen
- increase DC torque output as required to stay around 1750 rpm
- increase regen on AC controller
- repeat until maximum regen or maximum current on DC motor
- log each step and max current when limit is reached

The second test will have the AC motor shaft coupler bolted to a solid frame, and the AC controller will ramp up quickly (maybe 10v/ms) to about 3X rated slip (150 rpm), which should get me 4X rated current, or perhaps a bit more, but at a low voltage. My clamp-on meter is not real accurate, but this should determine if the LEM Hass current sensors are in the right ball park

[e*clipse]

I've made a bit of progress with the resolver stuff. Thank you very much, Dave, for the capacitance tips. Between tuning the circuit and adding some filter and powerbuffer capacitors, I've got a nice clean output. Oh, and another thing - I was looking into some timer IC's, and many only seem to run at much less than 5V - more like 3.6V. This actually turned out to be beneficial, because I was trying to run the system too close to the absolute limit of the op-amps. By reducing the input to 3.6V, they produced a much cleaner, less distorted output.

By the way, I was looking into the attiny IC you recommended for a 50kHz oscillator - are you referring to the ATtiny4 from Atmel? If this is it, it's basically a very basic microcontroller that could be programmed for a speed.

Another possibility, although not nearly as cheap would be the LTC6907 from Linear Technologies:
http://cds.linear.com/docs/en/datasheet/6907fa.pdf
It's about $3.00 in qty 1 from digikey, but it's a dedicated timer chip, designed for this job and identical to the ATtiny in size. All it needs is a precision resistor (about $1.00) to set the speed. Perhaps for the "don't need to mess with it" factor, the extra money is worth it.

Oh, yeah - another unexpected thing - my signal generator is a source of the noise. Simply generating square waves by definition creates a lot of noise. However, my circuit filters most of it, except for the spikes at 100kHz. Tuning for a specific "clean" speed like 62.5kHz may not be necessary, because the timer chip will have its own noise issues and because filtering out 100kHz is easy and already done.

I've also got the input circuit almost working. The first test produced a nice sinewave output at 2X the rotor rpm. The main issues right now are better synchronizing with the carrier frequency and cleaning up jittery peaks. The zero crossings are very cean.

[MPaulHolmes]

I'm using the attiny25. Nothing wrong with a dedicated chip (ltc6907)whose purpose is specifically for making pulses! By the way, I just programmed the attiny yesterday, and rediscovered that it is 62.5khz, not 50khz. Oh well. My memory was a bit faulty.

I found a minor mistake on the board. Q7 and Q8 need to be rotated 180 degrees before inserting them. I had fixed that a long time ago, installed a new version of designspark, and forgot that I had the old bad copy of that footprint. Oh well.

[e*clipse]

Hmmmm that's strange - I wonder what's up with 62.5kHz?? Seriously - why would they use that frequency?

Thanks for the heads - up about Q7 and Q8!

- E*clipse

Originally Posted by MPaulHolmes

I'm using the attiny25. Nothing wrong with a dedicated chip (ltc6907)whose purpose is specifically for making pulses! By the way, I just programmed the attiny yesterday, and rediscovered that it is 62.5khz, not 50khz. Oh well. My memory was a bit faulty.

I found a minor mistake on the board. Q7 and Q8 need to be rotated 180 degrees before inserting them. I had fixed that a long time ago, installed a new version of designspark, and forgot that I had the old bad copy of that footprint. Oh well.

[MPaulHolmes]

I tested the time the FOC code takes to execute in the interrupt. I have 3000 cycles available, and it took between 455 and 498. haha. Lots of room! And that was all with just C. I could probably get it down to like 490 with a few more optimizations, but why bother?! Microchip's code (with assembly language) took 1/3 of the 30MHz. Mine is 1/6. That's 2 times slower. am I missing something? LOL. I did whip out every trick in the book.